How long does it take to install fiber optic connectors? (April 16, '99)
How long does it take to install 6 fiber optic connectors? Half an hour? That's about the right time if you use the manufacturer's estimates for a typical adhesive type connector. It really doesn't seem to matter whether it's oven-cure expoxy, Hot-Melts (a 3M trademark) or anaerobics, they all take about the same time.
But how long does it REALLY Take? We'll here's a tale from the real world:
I helped a former student do a "simple" termination of a 6-fiber distribution cable in a old house recently given to Bradford College in Haverill, Massachusetts. The house was only 600 feet (just under 200 meters) from the telecom center for the campus. The fiber had been pulled already with innerduct in a 4 inch conduit.
Easy half-hour job, right? Hardly! From the time we parked the van to leaving for lunch was 2 hours and 15 mintues. But then, we didn't just terminate 6 fibers. First,we unloaded the van and brought everything inside. We set up our Fiber U toolbox and plugged in the portable connector curing oven to warmup while we got the cable ready. We had to run the innerduct across an open basement ceiling, cut it to length, cut the fiber optic cable to length, leaving lots of extra length for future moves or repairs. We had to mount a wall mount box to lock up the terminations. Then we were ready to do the connectors.
The connectors did indeed take 30 minutes. 100% yield. Very low loss.
The cables had to be tested, dressed, service loops tied up and the box locked. Then we had to clean up everything, pack all our tools, dump the garbage and load everything back in the van.
Yep, 6 connectors took 30 minutes. But the job took 2 hours and 15 minutes. Worth remembering when you estimate!
Now before you copper types go using this to knock fiber, how long do you think it would have taken to install 3 Cat 5 jacks under the same circumstances? Based on my experience in teaching copper installation for the last year, I'll bet on about 2 hours!
Report From OFC '99 (March 1, 99)
I spent last week at the OFC show in San Diego, and was quite frankly blown away by the activity level. Almost 10,000 attendees and exhibitors crowded into the Convention Center, filled hotels and busses and generally looked prosperous. Visitors were from all over the world. Makes me believe the industry is doing well.
As expected, every other booth was touting the latest DWDM technology. Like couplers ten to fifteen years ago, more suppliers are in the business than there are customers to buy the parts, so we'll see a bloodbath someday, when the accountants and Vulture Capitalists get wind of the real market.
Gigabit Ethernet had a good presence, as did the SFF connectors. I finally saw a MU connector in the flesh and it's CUTE! It really looks like "Honey, I shrunk the SC." See our RUMORs page for more stuff from OFC.
One interesting note: in the past, a few of the fiber optic market researchers had a tabletop exhibit or small booth, but this year Ryan, Hankin, Kent had a gigantic booth and strangely I couldn't find pioneer KMI anywhere.
Is the show worth the visit? Well if you live on the eastern part of the country, it will be in Baltimore next year, so come take a look. If you are an installer, don't travel too far, as it has little you can relate to. Fiber optic manufacturers should bus all their employees to the exhibits for a day, just to "wow them" with the technology.
What happens when we saturate the market? (March 1, 99)
A staff-written article in CED (a CATV magazine) last month talks about the time in the near future (5-7 years or so) when the CATV market is practically fully fibered. Unlike the long distance telco market, it's is not likely that we will see lots of alternative routing, competitive companies, etc. In other words, the fiber optic CATV suppliers either must come up with a killer upgrade or look elsewhere for business.
The article implies they will look overseas. But there is another scenario. If fiber, lasers and detectors are in oversupply, prices should fall, as has been happening recently in fiber itself. All of a sudden, the fiber to the home market might just start looking attractive to the suppliers and the telcos. Add to this the likely development of 1300 VCSELs and the holy grail of fiber optics might just be within reach!
Which Small Form Factor Connector Will Win? (February 3, 1999)
I am continually asked which of the small form factor connectors will win in the marketplace? I may be opinionated, but I am not omniscient! There are a few good bets you can make about them, however. These are:
The Lucent LC will survive. Well, Lucent is going to use it, aren't they? And they represent millions of connectors used themselves. Others will offer it as soon as they can get licensing set up, like Methode. The connector itself is not rocket science, but a solid evolutionary update of current, proven designs. It's compatible with known materials and termination technology. The smaller ferrule should actually have better performance and be easier to terminate. Guaranteed Winner!
Update (2/9/99) IBM Microelectronics will build transceivers for the Lucent LC. Supposedly they evaluated all the SFFs ( even the SC/DC, rumored to be an IBM inspired connector?) and chose the LC as the best.
The MU from NTT in Japan will survive for the same reason, but be second tier in volume. Same deal as the SC, limited supply in the beginning and higher price.
The SC/DC will be an IBM exclusive, like ESCON. Adding another alignment problem (rotation) makes it too touchy. It will be expensive - probably factory terminated - and rare.
The Opti-Jack, is, IMHO, the most simple/elegant design and the most versatile. I actually suggested this design at a POF meeting a year before it appeared, so I'm obviously prejudiced, but it takes proven technology and packages it perfectly. Panduit did their homework well. However, it's a tough market out there, and unless it has multivendor support, it will have a hard time surviving. It deserves more respect!(3-3-99, Maybe a 300,000 connector order for a fiber to the desk system will help establish their reputation.)
The MT-RJ is another elegant design, but flawed. Polishing the fibers is a trick, esp. between the molded-in pins on the male half. All the designs I have seen require a "cleave and leave" termination technique, and the cost of the connector itself will reflect the difficulty of making the connectors in the factory. The combination of the dual fiber design and the crummy cleavers that are usually shipped in "cleave and leave" kits will make field termination yield "iffy". It can work, but I suggest the field user buy a good ($1,000+) cleaver - remember you have to make TWO perfect cleaves! I've wondered about the ease of making transceivers also, with the fibers so close. With more companies behind this connector than any other, it will be a factor, but not without teething problems.
The Volition (a 3M trademark) appears to be a wonderful idea. Get rid of all the expensive hardware and create a simple v-block splice as a connector. Be willing to predict prices of 1/4th current prices. Address all the obvious problems like dirt getting into it properly and the availability of transceivers, back it by an advertising campaign big enough for the Super Bowl, give away a gazillion samples and train everybody interested. 3M has been doing it all right. The only negatives I see for it are the strength of the competition and the skepticism I hear from some experienced fiber types - they don't yet believe something so simple could work! (3-3-99, Getting into the business of supplying media converters, hubs and other networking hardware is another part of the 3M strategy. It will certainly help to sell a more complete solution.)
What would I design if I were in the connector biz ? I'd love to see an Opti-Jack with LC ferrules. The smaller ferrules could simplify transceiver design and probably reduce cost. Temination processes could be more flexible. Sure it's like a duplex LC, but looks more elegant. (Maybe mold a boot for the duplex LC instead?)
Manufacturers: Tell us what you think! (either connector or network equipment vendors)
Users (or potential users): Let us know your opinions.
Installers: Do you have any experience or intelligent guesses yet?
Your comments will be welcomed and posted.
Dialogue: (February 3, 1999)
Thanks to a Network Manager at a large university for this feedback. Let us know what YOU think!
The comment:
>
> first I want to thank you for all the wonderful information
on the fotec
> website. Next, I want to agree with one opinion and disagree
with
> another.
>
> Although I am not in any way biased toward copper (over fiber),
I think
> your scheme to save half the copper used in twisted pair
wiring is a
> little off base. The idea of installing 4-pair CAT5
> (5E, 6 - whatever) cabling is to make the copper cable plant
structured.
> You may well require only one pair for your telephone and
two pair for
> your ethernet connection, but gigabit ethernet has arrived
and it
> requires four pairs.
>
> Looking to the (near) future, you should be able to plug
your gigabit
> application into your phone outlet. No more phone vs data
vs fast data
> outlets. This may use up some copper, but will save tremendously
on
> installation, moves, adds and changes and cable management
over the life
> of the cable plant.
>
> Now for agreeing with your oppinion on "what is 'certified
fiber'
> anyway. We install a lot of backbone fiber, singlemode as
well as
> multimode, on the university campus. We do not get involved
with the
> installation of the fiber, but we get the test results. They
all read
> "pass".
>
> When we go to patch a fiber path, we use the OTDR to confirm
that the
> path makes it to the far end. At times I notice that the
dB losses are
> much higher than I would like them to be (I know- I shouldn't
use the
> OTDR to measure dB losses; still, it gives me an indication).
When I go
> to visually inspect the suspect connectors, I cringe at the
poor
> workmanship. If the contractors power meter gives a "pass"
reading with
> such workmanship, I fully agree that the standards are loose
and, in
> fact, encourages shoddy workmanship. However, it can be difficult
> convince someone with a fistful of "pass" test
results, that workmanship
> is a problem.
>
Thanks for your feedback. My somewhat "tongue in cheek"
condemnation of
UTP having 4 pairs is based on the fact that 99% of it will only
use 2
pairs since it will use Ehternet or Fast Ethernet on 2 pairs.
Very
little of it will ever be used for GBE, since that will be mainly
a
backbone technology, and most of the installed UTP, even Cat 5,
won't
pass Cat 5 specs! I got an independent confirmation of that recently
-
see http://tedmag.com/contract.htm
for an article quoting an anonymous
AMP source on Cat 5!
And I'm beginning to be a convert to OTDR testing of some premises
cabling - because of the number of bad connectors you can find
that way.
We get averages of 0.1 -0.3 dB from our students on their first
sessions
of termination. The 0.75 dB written into the EIA/TIA spec for
the
"cleave and leave" connectors with a splice in the backshell
makes it
easy to let terrible adhesive connectors pass!
Perhaps the Fiber Types are Using the Wrong Argument Against Copper Wiring! (12/15/98)
At a recent show, I had a booth next to the Copper Development Association. As trade associations go, these guys are super. They provide market information, ammo for competitive selling (esp. against fiber) and lots of educational material. The fiber industry is pretty wimpy in comparison!
But I picked up a fact book on copper and read it during slow times at the show. One statistic was missing: how much UTP cable (Cat 3 and Cat 5) is installed each year. So I asked the CDA representative.
His response, which I wrote on the back of the fact booklet, was not in cable length, but in pounds! They estimate that 100,000,000 pounds of copper were used for UTP cable manufacture last year! That's 50,000 TONS of copper!
To convert it to length, he gave me another fact: 4 pair UTP wire uses 10 pounds of copper per 1000 feet. That means that 10,000,000,000 feet (TEN BILLION!) of UTP was used last year.
Impressive, eh?
On the plane trip back home, I got thinking about this. 50,000 tons of copper! That's one heckuva lot of copper!
But wait! If that cable is used for telephones, only one pair of the wires are needed. And for most Ethernet links, only two pairs are used. Only a few Fast Ethernet links use all four pairs. So over half that copper is wasted!
That's right! Over 25,000 tons of copper each year are mined, smelted and drawn into wire, coated with some sophisticated insulation and jacketed, then pulled into a building NEVER TO BE USED AGAIN! And wire, by the way, always uses "virgin copper", not recycled like plumbing and many other uses.
Want to guess how much energy that takes? How much pollution is dumped into the air? How much non-renewable resources are wasted?
Now one could argue that we put in lots of dark fiber, and that takes resources and energy too. But does it compare to the waste with UTP?
Hey, CDA, want to offer a rebuttal?
More on Testing GBE Cable Plants (12/15/98)
We have completed a good portion of our R&D on cable plant loss with GBE VCSEL sources. Our comparisons with 850 nm LEDs used for lower speed networks (and our test equipment) shows fairly small variations. Take a look at the results.
Ignorance is No Excuse!
We're getting a whole new generation of companies involved with fiber and many are lacking in the depth of knowledge and understanding necessary to understand what’s going on in the fiber world. The standards groups aren’t helping either, as they deal in politics, not facts.I recently read two articles that were excellent examples of this problem. Both articles are based on the erroneous assumption that the margin for a network is calculated from EIA/TIA 568 cable specs!
The first article was from a tester company commenting that the typical LED source used for testing fiber loss would not be accurate for testing GBE loss because of the way it launches power. (We call that modal distribution, by the way, and here’s a good reference on it on the Fotec website.)
The article is superficially correct but missing at least one important point, maybe two. GBE will run on MM fiber with VCSELs but probably not without launch conditioning. The current fiber designs require a defocused launch to meet BW specs. The low numerical aperture of the VCSEL concentrates the light in the center of the fiber, which has index of refraction problems from the making of the preform. It causes all sorts of unpredictable BW performance. So the current intention is to make VCSELs that launch a defocused spot or even a "doughnut" filling about 2/3 the NA of the fiber - practically the same as a LED.
Thus the assumption that a connector with high loss when tested with a LED will have lower loss when used with a VCSEL may not be true. Furthermore, connector losses are practically never as bad as the 0.75 dB 568 allows! A ST or SC mated pair should be ~0.3 dB (our students routinely beat this in class) and the difference with a LED or laser will be much smaller than this article predicts, since the loss factors that are highly dependent on modal power distribution (lateral offset, angularity and air gap) are minimal.
Only the high loss connectors for which the 568 spec was loosened ( the cleave and leave connectors with a splice in the backshell of the connector ) will likely show big variations, and some of that effect may be coherence-caused interference.
We're starting a R&D project on this subject right now, but the VCSELs we have tested so far are not too good - unstable and seem to have lots of launch variability. We'll test them anyway and see what happens.
The other comment I have heard several times is that the 2.38 dB loss budget for GBE will be hard to test accurately. That appears to be the calculated loss of the cable plant using 568 component criteria and not the network margin!
Installers test with criteria much lower than 568 Appendix H. An instrument set to pass fail on 568 criteria will pass a cable plant with losses almost twice what it should be! These testers are going to help a lot of bad installs get "certified"! See my comments below!
Most cable people, it seems, are not familiar with network equipment! The transceivers that will be used by GBE have 7.5 to 15 dB dynamic range. (See http://www.sci.siemens.com/opto/fo/odl/gbt_trs.html for Siemen's specs as an example.) Transceiver dynamic range is the spec that needs to be considered for the loss the network can tolerate, not the 568 numbers for cable loss. With GBE, there are some power penalties associated with the performance of the cable plant, but they are worst case considerations. And most transmitters outputs are well above the minimum system spec.
Like every other fiber optic network, the system seems to be very conservatively specified!
The "Double Standard" of Standards
You can't sell a pencil in today's marketplace if it ain't standard,
it seems. We've got standards for everything under the sun. For
those of us in the communications cabling business, we have standards
for networks, network cabling, copper cabling, fiber optic cabling
and every single itty-bitty component that makes up any part of
these networks.
That's good, mostly. Standards insure interoperability among various
vendors, so the end user can install several different brands
of products and be sure they will work together. Large groups
of vendors have worked for days, months and even years in closed
rooms (with lawyers in the back making sure they don't get into
trouble with anti-trust laws) to write these standards.
But those of us who are married with children know that two people
can rarely agree on anything, and three or more means chaos! So
these standards, developed by dozens of people, end up being compromises,
often watered-down, and vague enough to allow marketing people
(not just lawyers) to find loopholes for a competitive advantage.
So while we have standards for network cabling, we have vendors
or groups of cooperating vendors offering a "warranty"
on their cabling only if you use their parts and their "certified"
installers. (Certified - now there's another overused word - but
we won't get into that just yet...)
How many of these vendors think their customers will be using
a network cabling plant for 15 to 20 years? How many of their
customers understand the warranty?
Manufacturer's warranties appear to really be marketing agreements
between companies and installers. You promise to install our components
and we'll help you find work and we'll guarantee the cable plant.
Installers are getting a bit upset at these programs. Did I say
A BIT? Seems they spend big money to get trained by the vendors
and get "certified" - but if they fail to meet sales
goals, they lose their certification - and their investment. Furthermore,
if they have a customer who wants another vendor that they are
not certified by, they lose a job.
The installers are really complaining to us at our Fiber U and Wire U seminars! They don't like feeling they are being pushed around by vendors. They don't like losing business because they can't afford to get certified by every vendor under the sun.
Several smart ones have told me they've overcome this issue with
customers with two questions: How long do you really expect
this cable plant to be able to support your network? And how much
extra do you think this 15-25 year (or longer) warranty will cost
you?
Does this make sense? Email your
comments! I promise to pass them along to the manufacturers!
The implications of the standard will reflect the points of view of the writers also. So network and cabling standards, written mostly by those who make 90% of their revenue from copper cable-based networks, reflect that prejudice. That's why EIA/TIA 568 limits the desktop connection to 100 meters for fiber and copper.
You don't think those copper guys on the committees want to handicap, fiber do you? Look at Gigabit Ethernet. They create a standard for GBE on copper that requires a Ph.D. from Bell Labs to tweak the cable to make it work, but spec fiber so conservatively it can't possibly fail if you ran it twice as far as the spec!
Paranoid? ME? Nope. Just experienced and cynical after 20 years
in fiber and 10 years in standards committees. I wish we could
make sure every end user knew where standards came from, that
they were voluntary interoperability standards written by the
vendors not laws handed down on stone tablets, and that those
who write those standards are neither objective, impartial, omniscient
nor infallible.
Send comments to Jim Hayes!
Speaking of Certification, what is a "Certified Fiber" Anyway???
Recently, we've seen the emergence of claims that instruments "certify" fiber optic cable installations to TIA, ISO and other industry specifications. What does all this mean?
By having the user input the number of connectors and splices in the cable plant, the instrument will calculate the loss according to EIA or ISO limits, compare it to the measured value and give a "PASS/FAIL" result. This is what seems to be meant by "certification."
Does cable plant "certification" mean the installation is good? Not at all. It means it meets EIA or ISO maximum limits, but that doesn't necessarily mean it's good. For example, EIA/TIA 568, Appendix H, specifies testing limits for fiber optic cable plants. It allows the following limits on component performance:
Fiber Loss: 3.75 dB/km @ 850 nm, 1.5 dB/km @ 1300 nm
Connectors: 0.75 dB
Splices: 0.3 dB
These values are a result of a "committee" decision for worst case installation. (See Double Standards above.) In today's world, typical field-installed ST or SC connectors will have losses averaging 0.3 dB or better. Fiber loss specifications are listed by many cable manufacturers as typically 3.0 dB at 850 nm and 0.7 dB @ 1300 nm.
Lets look at what a "PASS" test result based on EIA
limits can tell you. (In the table, "EIA/typical" means
losses calculated for EIA limits vs. typical limits.) Here's our
projected loss budget for two links.
|
|
|
|
|
|
|
|
| Example 1, horizontal, 90m, 2 conn | ||||||
| 850 | 0.34/0.27 | 1.50/0.60 | 1.84 | 0.87 | 0.97 | |
| 1300 | 0.14/0.06 | 1.50/0.60 | 1.64 | 0.66 | 0.98 | |
| Example 2, backbone, 2000m, 6 conn | ||||||
| 850 | 7.50/6.00 | 4.50/1.80 | 12.00 | 7.80 | 4.20 | |
| 1300 | 3.00/1.40 | 4.50/1.80 | 7.50 | 3.20 | 4.30 | |
Example 1
|---------------------------------------| 90 meters
On a "horizontal" run of 90 meters, these limits
will pass a fiber link that has about 1 dB more loss than it will
typically have. That might not be bad, unless one of the two connectors
is the culprit. A bad connector on either end will sneak through
this test.
Example 2
|-----|--------------|-----------/ /----------|------------|-----|
2000 meters
On a "backbone" cable, the results can be incredibly deceptive. A "PASS" link can have 4.2-4.3 dB more loss than a good link should typically have. This "PASS" result could have a link with 1 dB connectors throughout or a number of 0.3 dB connectors and one >2 dB connector! It could also easily mask another problem a human would not overlook. If the cable has been overstressed in installation, the loss of the fiber will be higher at 1300 nm due to bending losses, sometimes as high as the loss at 850 nm.
This "PASS/FAIL" approach with such loose limits would overlook major faults in installation because of the high acceptable connector and fiber losses. It might not be a problem in the beginning, but a bad connector may become worse over time or be unacceptable when cables are patched during moves or changes. Stresses in fiber can cause long term failure.
And what will happen when testing a link for Gigabit Ethernet with a 2.5 dB margin?! The standards people better get on this one now!
The problem is not only the makers of this instrument, its also the writers of the EIA/TIA standards. By using such loose specifications, they have placed the burden on the installer to make certain that certain the cables are installed properly and tested with some intelligence.
The only way to "certify" a fiber optic link properly is to tighten specs to a more realistic level and test individual components when test results indicate problems. On long distance links, telcos and CATV operators use OTDRs to check each connection, but these instruments are not designed for the short lengths of cable typical of LANs.
Another way to "certify" a link is to test it also against the margin of the network it is designed to support. But bad workmanship should not be passed just because the network will run on that loss!
*It's MY OPINION!
I've always been outspoken, never humble, often cynical, thoroughly opinionated and love debating issues. This webpage is purely my opinion, based on what I see going on in our industry. It does not represent an official position of my employer or any other employee. Just me!
I welcome - in fact I created this page to encourage - discussion. I welcome your comments, feedback, opinions, etc. I only ask you be civil and courteous. Discussion and debate are gentlemanly (or if you want to be PC, gentlepersonly) pursuits; making a contrarian point requires you provide a reasonable argument.
To insure the discussion remains genteel (free from vulgarity
or rudeness, polite), you must address your comments to me: email Jim Hayes.
Enjoy. And participate.
© 1998-2001, Jim Hayes
Return to the Fotec Home Page and check out the Cable U Page
Return to the Fotec Home Page and check out the new Cable U Page